The Queue Continuum: Applied Queuing Theory
summary
Star Trek framing, then concrete claims and knobs: Little's law shows service time times arrival rate gives offered traffic (example: six beamed every five seconds → offered traffic 41.5), utilization above ~90% makes queuing latency explode, and adding parallelism (more transporters/processes) or tuned concurrency (more threads) is quantified. Uses Brendan Gregg’s use method (utilization, saturation, errors), benchmarks Sidekiq concurrency history (30 → 10) and practical Puma per‑process queues, plus IO‑weight math to pick concurrency.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 2:30Queuing theory meets Star Trek Transporters
- 5:05Little's law explained
- 8:20USE method for capacity planning
- 11:34Parallelism versus concurrency
- 15:47Vertical and horizontal scaling
- 17:14Global VM Lock and Ruby threads
- 20:44Benchmarking concurrency
- 23:57Distributed job queuing with Sidekiq
- 26:43Web request queuing and Heroku routing
- 28:09System-wide queuing balance
- 29:11Closing remarks
related talks
38mA Practical Look at Performance Theory
40mResilient Design 101: Queue Theory
41mGetting Out of Quicksand, With DevOps!
31mWhat does "high priority" mean? The secret to happy queues
40mServiceability under high demand
36mFrom Request To Response And Everything In Between
36mFrom Request To Response And Everything In Between
31mFrom Bottleneck To Breakthrough: Conquering Applications Star...